app.py

from fastapi import FastAPI, Header, HTTPException
from pydantic import BaseModel
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
from typing import Optional

# -----------------------
# App
# -----------------------
app = FastAPI()

# 🔐 API KEY (keep same)
API_KEY = "sk-tinyllm-9f3a2c7e8b4d1a6c0e52f91d"

# ✅ Lightweight CPU model (NLP engine only)
MODEL_NAME = "Qwen/Qwen1.5-0.5B-Chat"

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    dtype=torch.float32
)
model.eval()

# -----------------------
# Request schema
# -----------------------
class Prompt(BaseModel):
    message: str

# -----------------------
# API key verification
# -----------------------
def check_api_key(authorization: Optional[str]):
    if authorization is None:
        raise HTTPException(status_code=401, detail="Missing API key")

    if not authorization.startswith("Bearer "):
        raise HTTPException(status_code=401, detail="Invalid API key format")

    token = authorization.replace("Bearer ", "").strip()
    if token != API_KEY:
        raise HTTPException(status_code=401, detail="Invalid API key")

# -----------------------
# Health check
# -----------------------
@app.get("/")
def root():
    return {"status": "TinyLLM RAG NLP API running"}

# -----------------------
# Chat endpoint (RAG-safe)
# -----------------------
@app.post("/chat")
def chat(
    prompt: Prompt,
    authorization: Optional[str] = Header(None)
):
    check_api_key(authorization)

    # 🚫 IMPORTANT:
    # DO NOT inject system identity here.
    # Your RAG prompt already contains ALL rules.
    messages = [
        {
            "role": "user",
            "content": prompt.message
        }
    ]

    input_ids = tokenizer.apply_chat_template(
        messages,
        return_tensors="pt",
        add_generation_prompt=True
    )

    with torch.no_grad():
        output_ids = model.generate(
            input_ids,
            max_new_tokens=220,        # controlled output
            temperature=0.0,           # 🔥 NO hallucination
            top_p=0.7,
            top_k=20,
            do_sample=False,           # deterministic
            repetition_penalty=1.1,
            eos_token_id=tokenizer.eos_token_id
        )

    response = tokenizer.decode(
        output_ids[0][input_ids.shape[-1]:],
        skip_special_tokens=True
    ).strip()

    return {
        "response": response
    }